Menu App

Workflow

1. Locate and Read the Menu

   • IF: $ARGUMENTS contains a file path → read that file
   • IF: $ARGUMENTS is empty → prompt for the menu file path
   • Supported formats: PDF (Read tool), images/PNG/JPG (Read tool), HTML/MD/TXT (Read tool), DOCX (extract via Bash textutil on macOS or similar)
   • IF: file not found or unreadable → stop and report
   • Example: /menu-app ~/menus/pizza-palace.pdf → Read the PDF, see the menu pages
   • Tool: Read (for PDF, images, text), Bash (for DOCX conversion if needed)

2. Extract Structured Data

   • Parse the menu content into structured data. Extract ALL of the following:
   • Restaurant info: name, tagline/motto, phone number(s) with contact names, email, address, delivery policy, business hours, any special notes
   • Categories: section headings that group menu items (e.g., "Appetizers", "Salads", "Main Courses", "Desserts", "Drinks")
   • Menu items — for each item extract:
     • Name (in original language)
     • English name/translation (if menu is non-English, translate; if already English, skip)
     • Price and unit (per lb, per piece, per kg, per liter, etc.)
     • Variants/sub-items (e.g., fillings, sizes, cooking options) with their individual prices if different
     • Minimum order constraints (e.g., "min 5 of same filling")
     • Seasonal or availability notes
     • Brief description if present on the menu
   • Auto-tag items based on ingredient keywords in the name/description:
     • meat — beef, pork, lamb, salo, liver, tongue
     • chicken — chicken, poultry, turkey, duck
     • fish — fish, salmon, tuna, shrimp, crab, herring, sturgeon, seafood
     • veg — items with no meat/chicken/fish indicators
     • popular — if the menu highlights it, or if it's a well-known signature dish
     • seasonal — if marked seasonal or limited availability
   • IF: price is "by arrangement" or missing → set price to null, display "by arrangement"
   • IF: item has variants at different prices → create separate entries OR a single entry with variant details
   • Example: "Блинчики с начинкой: мясо/творог/яблоки — $3/шт, мин. 5" → name: "Blini with filling", price: 3, unit: "pc", details: "meat / cheese / apple / potato / cabbage / mushroom / cherry — min 5 of same filling", tags: []
   • Example: "Fish & Chips — $14.99" → name: "Fish & Chips", price: 14.99, unit: "", tags: ["fish"]

3. Determine Branding

   • Pick a color scheme that fits the restaurant's vibe:
     • IF: menu has visible brand colors → match them
     • IF: upscale/fine dining → muted, elegant palette (dark backgrounds, gold/cream accents)
     • IF: casual/family → warm, inviting palette (earthy tones, readable)
     • IF: fast food/takeout → bold, high-contrast palette
     • IF: no strong signals → use a warm neutral palette
   • Choose an accent color for buttons, highlights, and the cart
   • Ensure sufficient contrast for accessibility (text on background)
   • Example: Russian catering with yellow-toned PDF → warm brown/sienna accent on cream background

4. Generate the HTML App

   • Read and follow ./cookbook/html-app-spec.md for the complete HTML generation spec
   • Write a single self-contained HTML file — all CSS and JS inline, zero external dependencies
   • The spec covers: header with contacts, search/filters, menu card grid, cart system with localStorage, order text export, call/navigate buttons
   • Example: 90-item Russian catering menu → single HTML file with 5 category sections, filter pills for meat/fish/veg/chicken, slide-out cart, copy-to-clipboard order text
   • Tool: Write

5. Determine Output Path and Write

   • IF: $ARGUMENTS contains a second argument → use as output path
   • ELSE: derive from restaurant name → e.g., pizza_palace_menu.html in the current directory
   • Write the complete HTML file
   • Example: /menu-app ~/menus/sushi-zen.pdf → writes sushi_zen_menu.html
   • Example: /menu-app ~/menus/cafe.pdf ~/Desktop/cafe.html → writes to specified path
   • Tool: Write

6. Verify

   • Read the generated file to confirm it's well-formed
   • Report to user: output path, number of items extracted, number of categories, any items with missing prices or ambiguous data
   • IF: items were ambiguous or skipped → list them so the user can review
   • Example: "Generated pizza_palace_menu.html — 47 items across 5 categories. 2 items marked 'price by arrangement'."
   • Tool: Read